Airport Passengers Movement Trend Prediction by Means of Machine Learning and Deep Learning

In this work, machine learning algorithms known as SARIMA and deep learning are used to build passenger movement trend prediction models. The constructed models are used to forecast the cumulative number of passengers arriving and departing from the airport daily using data from Minangkabau International Airport, Indonesia. Several deep learning architectures and hyper -parameters are iterated to select the best model in terms of accuracy and computational time. The evaluated deep learning architectures include dense, convolution and LSTM models. Based on the conducted study, all models have comparable accuracy with near 10% mean average percentage error yet with diverse computational times. The most accurate model for this case is found to be the convolutional deep learning model, which has 9.01% mean absolute percentage error and 11.48 seconds of computational time. On the other hand, the fastest prediction model is SARIMA with 9.19% mean absolute percentage error and 2.96 seconds total computational time.